124884 - Moz. 0.9.7: Apparent infinite looping as libgklayout.so calls NSGetModule()

Status: VERIFIED WORKSFORME

(SeaMonkey :: General, defect)

Description

[Steven Augart]

Mozilla 0.9.7 apparently repeatedly loops at a stack level 80 calls deep;
most of the calls are into NSGetModule.

Symptoms are: Mozilla stops updating its X windows, and an instance of
mozilla-bin begins eating all available CPU.

I had this happen while I was visiting EBAY.COM.

Unfortunately, I was not able to extract the exact URL I was visiting from the
browser image; when I restarted the browser, my browser history for that day was
missing.

I am willing to attempt to find that URL by experimentation if necessary.

I attached gdb to the mozilla-bin process and took repeated snapshots of the 
stack, letting it execute a bit and then snapshotting again.  This gave me
various addresses for the instruction pointer and some slightly-varying
backtraces.  I'll attach these.

Comment 1

[Steven Augart]

Attachment: A GDB backtrace of mozilla stopped while chewing up CPU.

Comment 2

[Steven Augart]

Attachment: Second backtrace of mozilla in the act of chewing up CPU; different instruction pointer.

Comment 3

[Steven Augart]

Attachment: A third backtrace (abbreviated) of mozilla in the act

Comment 4

[Steven Augart]

Attachment: A collection of abbreviated backtraces and instruction pointer values, as I repeatedly interrupted and continued Mozilla.

I apologize that I don't have better data here.  I would be glad to
download a Linux build that already had a full set of debugging symbols and
source files attached to it.  I assume you don't have such a thing; maybe I
should start building one for myself, and offer it to share?

Comment 5

[Karl Johan Kleist]

Excellent bug report. However, you may want to set the severity level to
"critical", and add the keyword "hang".

Comment 6

[Matthias Versen [:Matti]]

Reporter: Can you please upgrade to 0.9.8 ? 

Comment 7

[Matthias Versen [:Matti]]

wfm - no response

Comment 8

[Steven Augart]

Matti:  This does now work fine for me with later Mozilla builds.  I think we
can assume that the bug was fixed in later releases.  WFM seems like a fine
resolution to the problem.

(I'm going through my old bug reports and trying to clean them up.  Sorry for
the delay.  Your email asking me to report on my experience with a later build
must've come on a heavy spam day and gotten lost in the fray.)

Comment 9

[Matthias Versen [:Matti]]

thanks :-)